370 likes | 482 Vues
This thesis explores the automation of CAD modeling through Knowledge-Based Engineering (KBE) methods. It addresses the industry's growing adoption of KBE, particularly among late adopters facing slow change and bureaucratic challenges. The research focuses on conducting a cost-benefit analysis to determine if the resource savings from CAD automation justify its development costs. Interviews with industry professionals reveal significant potential for improved efficiency, reduced modeling inconsistencies, and better design iterations, thereby promoting wider KBE adoption in engineering sectors.
E N D
A CASE STUDY IN KNOWLEDGE-BASED CAD MODEL AUTOMATION Thesis Defense Andrew Lowe
Background • What is Knowledge-Based Engineering? • Growing industry adoption • Late Adopters • Large bureaucracies, slow change • No direct contribution to revenue stream
Background and Research Question • What is Knowledge-Based Engineering? • Growing industry adoption • Late Adopters • Large bureaucracies, slow change • No direct contribution to revenue stream • Does a cost-benefit analysis of CAD modeling automation yield a high enough resource savings relative to the resources invested in the application’s development to justify its use in industry?
Significance • Great potential for saved resources in partner company • Possible impetus for greater KBE focus in partner company • Possible impetus for more widespread KBE adoption in similar companies. • Cost of automation relative to benefits not quantified
Literature Summary • KBE • Basic principles, industry implementation • No discussion of financial savings or cost • CBA • Basic formula, breakdown of formula • Identification/estimation of intangible benefits • Eliciting expert knowledge • Complex Surfaces • Mathematical basis and troubleshooting errors
Research Type • Two phase project • Development of application • Analysis of implementation • Resource savings • Surveys/Interviews • Cost-Benefit Analysis
Development Methodology • Project specs already in place • Solid creation/validation • Additional features • Cyclic design/validation scheme • Development time recorded
CBA Methodology • Once effected areas are identified, benefit to those areas must be quantified • More use of interviews and questionnaires • Many tangential effects will in intangible in the context of CBA • Sum total benefit and total costs
Progamming: GUI • Overarching goal: Simplicity and supportability • Inputs grouped into required/optional file input and text input • Optional groups collapsed by default • Execution buttons not active until required values entered
Programming: File Parsing • ASCII files with Cartesian points • Organization different for each file type • Blade file • Surface normal file • Flowpath file/spline reference file • Differing format forces additional complexity in code • Only similarity in code for opening file • Loops for organizing data into usable format vastly different with little code reuse
Programming: Geometry • Three geometry types • Blade/Splitter • Surface Normals • Flowpath/Spline reference • Blade Geometry • Input points organized by face • Splines made for forward three faces, trailing edge • Surfaces made from splines • Surfaces knit into solid • Individual face surfaces made for FEA integration
Geometry (cont’d) • Surface normals • Points organized by pressure side/suction side • Subdivided by hub/tip • Splines made between hub/tip on each side • Flowpath/Spline reference • Simple list of points • Spline made from points
Programming: Ancillary Features • Variety of text and numeric embedded data • Simple coding • Similar to variable declaration
Programming: Documentation • Fully-commented code • More in-depth than typical commenting • Targeted toward novice users • Detailed flowchart • Short overview/introduction document
Data Collection: Participant Identification • Initial interview PLM project manager • Overview of impeller design process • Identified where application would affect process • Areas finalized in follow-up interview • Prospective participants approached by company point of contact • Positive respondents forwarded to researcher • Final participants: 3 modelers, 5 engineers, 1 analysis integration expert, 1 project manager
Data Collection: Interviews • With one exception, interviews conducted over the phone • Interviews started with explanation of application and a few basic questions • How do you see this affecting your area? • How would you use the application? • How would the application save or cost you time? • Additional questions based off their responses • Follow up interviews conducted after receiving new information from other participants
Modeler Interviews • Focused mainly on modeling consistency • Large amounts of wasted time troubleshooting unfamiliar models • Get called away from current work to troubleshoot their own models • Only about 1 hour per iteration in direct labor savings • Much larger time savings from not having to reconstruct adjacent geometry with each iteration
Engineer Interviews • Reduced idle/wait time • More process control • More design iterations • 3-5 total currently, 15-20 possible in one night with automation • Higher quality • More in-spec options for customers • Confirmation of modelers modeling consistency concerns
Analyst Interviews • Analysis integration process • Automated meshing • Allows for batch modeling/FEA runs • Time savings data for both • Provided proprietary document confirming and expounding upon their points • Confirmed modeler/engineer concerns and data
Project Manager Interviews • Wide-view design process information • Participant identification • Go-between with high-level non-participants • Employment cost data • Other financial data • New information for follow-up interviews
Data Analysis: CBA • Total Benefit = Gross Benefit – Gross cost • Gross Benefit = Quantifiable Benefit + Intangible benefit (A) • Gross Cost = Quantifiable Cost + Intangible Cost (A)
CBA: Cost • Four ways of interpreting quantifiable cost • Interviews yielded no information hinting at additional real or intangible costs
CBA: Benefit • Three ways of evaluating benefit: • Current process with dumb solid macro • Frequently-used process with spline creation automated • Completely manual process • Numbers calculated using time savings estimates from participants and company’s $100/hr average cost of employment
CBA: Intangible Benefits • Many intangibles lacked data for proper estimation • Lower error cost from modeling inconsistency • Higher quality/more customer options from multiple iterations • Lower operating costs running batch at night • Decreased time to market
Intangible Benefits (cont’d) • Decrease in overall project time from elimination of engineering/FEA/modeling bottlenecks • Possibly greater than two weeks • Estimated based on publicly available defense contracts • 40% share in $2.4 billion, 6.5 year project • IDA estimated 11% profit margin • $356,000 per-day savings • $3.56 million two-week savings
CBA: Results • Analysis without intangible benefits:
CBA Results (cont’d) • Most cost-effective development method: in-house employee • For benefit to be seen from direct labor savings, must be used in batch or over multiple projects • Chart provides baseline minimum benefit, as not enough data could be gathered to support intangible benefits as a guaranteed non-zero value
CBA Results (cont’d) • Intangible benefits potentially dwarf quantifiable benefits • Application has high potential to provide net benefit • Could pay for most expensive development cost in a few hours of reduced project time • Time savings over three days delivers an order of magnitude ROI on its own • Does not take into account the other non-estimable intangibles. • Due to lack of information, intangibles cannot be assumed to be non-zero, should be stated in a 0-maximum format
Conlusions • Does CAD automation show a high enough cost/benefit ratio to merit wider industry implementation? • Yes • Tangible cost data was for single project • Re-use over multiple projects compounds savings and provides net benefit from direct labor savings alone
Conclusions (cont’d) • Factoring in intangible benefits, total savings becomes massive • Potential order-of magnitude ROI • These savings compound over multiple projects as well, potentially reaching two orders of magnitude ROI • Thesis evaluated a single automation routine for a fraction of a single part, use of automation on multiple parts could increase potential benefits even higher.
A CASE STUDY IN KNOWLEDGE-BASED CAD MODEL AUTOMATION Thesis Defense Andrew Lowe